iDQ: Statistical Inference of Non-Gaussian Noise with Auxiliary Degrees of Freedom in Gravitational-Wave Detectors

TL;DR

iDQ is a real-time supervised learning framework that detects non-Gaussian noise artifacts in gravitational-wave detectors using auxiliary data, improving data quality assessment and noise identification in gravitational-wave astronomy.

Contribution

This paper introduces iDQ, a novel statistical and machine learning framework that autonomously detects non-Gaussian noise in gravitational-wave data using auxiliary channels, operational in real-time at LIGO.

Findings

iDQ successfully reproduces known data quality monitors.

iDQ identifies noise artifacts not flagged by existing methods.

iDQ operates in low latency during the advanced LIGO era.

Abstract

Gravitational-wave detectors are exquisitely sensitive instruments and routinely enable ground-breaking observations of novel astronomical phenomena. However, they also witness non-stationary, non-Gaussian noise that can be mistaken for astrophysical sources, lower detection confidence, or simply complicate the extraction of signal parameters from noisy data. To address this, we present iDQ, a supervised learning framework to autonomously detect noise artifacts in gravitational-wave detectors based only on auxiliary degrees of freedom insensitive to gravitational waves. iDQ has operated in low latency throughout the advanced detector era at each of the two LIGO interferometers, providing invaluable data quality information about each detection to date in real-time. We document the algorithm, describing the statistical framework and possible applications within gravitational-wave searches. In particular, we construct a likelihood-ratio test that simultaneously accounts for the presence of non-Gaussian noise artifacts and utilizes information from both the observed gravitational-wave strain signal and thousands of auxiliary degrees of freedom. We also present several examples of iDQ's performance with modern interferometers, showing iDQ's ability to autonomously reproduce known data quality monitors and identify noise artifacts not flagged by other analyses.